TDX 2026 — Key Takeaways
- Salesforce announced Headless 360 at TDX 2026 — every platform capability is now exposed as an API, MCP tool, or CLI command so AI agents can operate Salesforce without a browser.
- 100+ new tools and skills ship today, including 60+ MCP tools, 30+ coding skills, native React on platform, and Agentforce Vibes 2.0 with multi-model support (Claude Sonnet, GPT-5).
- A new Agentforce Experience Layer renders rich, interactive agent components natively across Slack, Teams, ChatGPT, Claude, Gemini, mobile, and WhatsApp — build once, deploy anywhere.
- New Testing Center, Custom Scoring Evals, Agent Script (open-sourced), Observability, Session Tracing, A/B Testing, and Agent Fabric address the hardest part of agentic enterprise: trust and governance at scale.
- Pricing is shifting from per-seat to consumption-based for Agentforce — a structural change every CIO and procurement leader must plan for.
What Is Salesforce Headless 360?
For 25 years, "using Salesforce" meant working inside Salesforce — consoles, tabs, page layouts, clicks. Headless 360 inverts that model. Every capability across Data 360, Customer 360, Agentforce, and Slack is now programmatically callable, so humans and autonomous agents can build, act, and deliver experiences on any surface.
Salesforce co-founder Parker Harris framed it bluntly: "Why should you ever log into Salesforce again?" Headless 360 is the company's answer — the platform stays; the UI becomes optional.
The four-layer architecture, now fully programmable
System of Context
Unified, real-time business data agents can act on
System of Work
Decades of business logic and workflows, now agent-orchestrated
System of Agency
Build, deploy, govern agents at scale across every channel
System of Engagement
Where humans and agents collaborate to get work done
What It Means for Customers
Customers stop being constrained to Salesforce's native UI. The capabilities they paid for can now show up wherever their employees and end customers already work — and agents can do the heavy lifting between them.
Faster time-to-value
Engine deployed a customer-service agent in 12 days that now resolves 50% of cases autonomously — the kind of velocity that used to take quarters.
Work happens where the conversation lives
Approvals, decisions, rich data, and full workflows now render natively inside Slack, Teams, WhatsApp, and even ChatGPT/Claude — no UI handoff.
Existing investment, multiplied
Years of Salesforce data, business rules, sharing models, and compliance controls become the substrate for every new agent. No re-platforming.
New economics
Consumption-based Agentforce pricing rewards efficiency. Outcomes — not seats — define spend.
Channel reach without channel projects
The Experience Layer means one agent definition deploys to six surfaces. That used to be six separate projects.
What It Means for Developers
For the first time, Salesforce has handed developers a true build-anywherecontract. The platform is now usable from any IDE, any coding agent, and any front-end framework.
60+ MCP tools, 30+ coding skills
Direct, live access to your org from Claude Code, Cursor, Codex, Windsurf — no more switching to the Salesforce IDE.
Agentforce Vibes 2.0
Native dev environment with full org awareness, multi-model (Claude Sonnet, GPT-5), open agent harness for Anthropic + OpenAI SDKs.
Native React on platform
Build fully custom front-ends in React, connect to org metadata via GraphQL, inherit all platform security.
DevOps Center MCP + Natural Language DevOps
Describe what to deploy; the agent runs the CI/CD pipeline. Salesforce reports up to 40% faster cycle times.
Agent Script (open-sourced)
A deterministic DSL that combines explicit business logic with LLM reasoning — solving the "brittle agent" problem.
Testing Center + Scoring Evals
Score whether an agent made the right decision, not just whether it executed. A genuine engineering discipline for non-deterministic systems.
What It Means for System Integrators
Headless 360 is, in many ways, an SI's release. The work shifts from config-and-clicks to architecture, agent design, evaluation engineering, and lifecycle governance — exactly the disciplines mature partners are built around.
New service line: Agent Lifecycle Management
Testing Center setup, custom scoring evals, observability dashboards, A/B promotion strategy, Agent Fabric governance — these are billable, recurring, expert services.
Agent Script as a delivery artifact
Versioned, auditable, open-source DSL for agent behaviour. SIs can package proven Agent Script templates per industry — much like managed packages, but for agents.
Multi-channel by design
One agent, six surfaces. SIs that previously delivered separate Slack, WhatsApp, mobile, and portal projects now deliver an Experience Layer once and orchestrate across channels.
AgentExchange go-to-market
10,000 Salesforce apps, 2,600+ Slack apps, 1,000+ Agentforce agents — plus a $50M Builders Fund. SIs with productised IP have a clear monetisation path.
Adoption & Enhancement is the new norm
Probabilistic systems drift. Continuous tuning, eval, and re-deployment make ongoing AES engagements structurally essential — not optional.
Consumption pricing changes the conversation
Optimising agent runs, prompts, model selection, and tool usage is a new FinOps discipline SIs can own.
Impact on the Ecosystem
From walled garden to open substrate. Salesforce is integrating openly with OpenAI, Anthropic, Google Gemini, Meta LLaMA, and Mistral, and supporting multiple agent SDKs. The platform stops competing with the LLM layer and positions itself as the place where context, workflow, trust, and engagement converge.
MCP, CLI, API — all three. Rather than betting on a single protocol, Salesforce exposes every capability across all three access patterns. Even Salesforce's own EVP told VentureBeat the team is "not at all sure" MCP will remain the standard. Headless 360 is deliberately protocol-agnostic.
A direct response to "vibe-coded CRMs." Some companies are now attempting to replace CRM with bespoke LLM-generated apps. Headless 360 counter-positions Salesforce as the most agent-friendly substrate available — keep the data, workflows, and trust; bring any agent.
ISV and partner economics shift. AgentExchange listings already show measurable impact: Notion cut sales cycles from four months to three weeks; Docusign processed 200+ private offers in Q4'25 with 60% faster signature time. The marketplace becomes a primary distribution channel for agentic IP.
SaaS pricing is being rewritten in public. Per-seat licensing assumes humans are the unit of work. As agents take over execution, consumption-based pricing isn't a Salesforce experiment — it's the new ceiling the entire enterprise software industry will be measured against.
KVP Observations & Assessment
1. This is the most consequential Salesforce announcement since Lightning.
Lightning re-platformed the UI. Headless 360 makes the UI optional. The commercial implications — channels, pricing, partner roles — are larger than any release we have analysed in the last decade.
2. The trust layer is the real moat.
Coding agents can call any API. What they cannot generate from a blank prompt are years of permissions, sharing rules, audit trails, escalation policies, and SLA logic. That accumulated trust layer is what keeps Salesforce relevant in the agentic era — and it is now exposed natively to every agent your enterprise runs.
3. Most enterprises are not ready for probabilistic software.
Testing Center, Scoring Evals, Observability, Session Tracing, A/B Testing — these are not nice-to-haves. They are the difference between a pilot agent and a production agent. We expect agent QA and evaluation to become a standalone competency inside CIO organisations within 12 months.
4. Slack just became strategic infrastructure.
Custom AI agents on Slack have grown 300% since January. With the Experience Layer rendering rich workflows inside Slack, the conversation truly becomes the interface. Customers without a Slack adoption strategy will lose ground on the agent-collaboration axis.
5. Consumption pricing demands FinOps discipline now.
We strongly recommend customers pilot Agentforce with explicit budget guard rails, prompt-cost monitoring, and model-routing strategies from day one. Without it, the economics that look attractive in pilots will surprise you in production.
6. The risks worth naming.
- Protocol churn: MCP may be displaced. Build with the abstraction (Headless 360 covers API/CLI/MCP), not the protocol.
- Agent sprawl: Without Agent Fabric governance, multi-vendor agent estates become ungovernable inside 18 months.
- Skill gap: The new disciplines — eval engineering, agent observability, prompt FinOps — are scarce. Plan now.
7. KVP's recommended next steps for customers
- Run a 6-week Headless 360 readiness assessment on your top 3 use cases.
- Pilot one customer-facing agent (deterministic, Agent Script driven) and one employee-facing Ralph-Wiggum-loop agent in parallel — they teach different lessons.
- Stand up Testing Center + Custom Scoring Evals before scaling — eval-first, not deploy-first.
- Establish consumption budgets and FinOps reporting at the agent level, not the org level.
- Identify two channels (e.g., Slack + WhatsApp, or Teams + ChatGPT) and use the Experience Layer to deploy a single agent across both as a proof point.
Bottom line
Headless 360 is Salesforce dismantling its own front door so every agent in the world can walk through it — and still be governed, auditable, and connected to the data and workflows that took enterprises decades to build. For customers, developers, and SIs willing to embrace eval-first, agent-native delivery, this is the largest expansion of platform surface area in the last 10 years.
Plan your Headless 360 strategy with KVP
Talk to our Agentforce architects about a readiness assessment, eval framework, or pilot design.
Sources: Salesforce Newsroom — "Introducing Salesforce Headless 360" (April 2026); VentureBeat — "Salesforce launches Headless 360 to turn its entire platform into infrastructure for AI agents" (April 2026).